Mimicking oxidative damage in γS-crystallin with site-specific incorporation of 5-hydroxytryptophan

Abstract

The human eye lens plays an essential role in vision by focusing light onto the retina. This transparent tissue consists of densely packed crystallin proteins that exhibit remarkable solubility despite minimal protein turnover. Post-translational modifications that accumulate over a lifetime can reduce crystallin solubility, resulting in the precipitation or phase separation of protein aggregates. Oxidation is a common type of modification that can cause such opacification of the lens, particularly in age-related cataract. Here, we study the oxidation of W163 in γS-crystallin, a structural lens protein that is particularly vulnerable to oxidative stress. We were motivated by previous findings reporting the oxidation of this residue in diseased and UV- and γ-irradiated samples. Using genetic code expansion (GCE), we incorporated an oxidation mimic, 5-hydroxytryptophan (5HTP), at position 163 of γS-crystallin (γS-W163(5HTP)). This subtle change in the structural and electronic properties of its side chain is hypothesized to destabilize the hydrophobic core of the C-terminal domain. γS-W163(5HTP) was characterized and compared to the wild-type (γS-WT). Although the overall fold and stability of the two proteins were comparable, the aggregation of γS-W163(5HTP) was triggered at notably lower temperatures compared to γS-WT. Subsequent investigation of this observation using both simulations and experiments suggests a potential mechanism for polymerization as well as oxidation-induced conformational changes that may cause susceptibility to thermal aggregation. Our findings highlight the utility of GCE platforms for systematically evaluating the impact of post-translational modifications on disease-related proteins.

Graphical abstract

Why it matters

The human eye lens is composed primarily of very stable, long-lived proteins called crystallins. Unlike most proteins, which are continuously recycled, crystallins must remain stable and soluble throughout the human lifespan. Aging causes damage to the lens, primarily via photochemical oxidation. Over time, this causes crystallin aggregation and leads to cataract. Although understanding oxidative damage is critical to understanding cataract formation and how it can be prevented, it is difficult to study in native biological systems. Here, we use genetic code expansion to introduce an oxidation product, 5-hydroxytryptophan (5HTP), in a key site in human γS-crystallin, enabling it to be specifically investigated under controlled conditions. Replacing a critical tryptophan residue with 5HTP leads to reduced stability and increased aggregation.

Introduction

The human eye lens comprises layers of transparent fiber cells that lose their nuclei and other organelles during early development. What is left behind is a highly refractive hydrogel, populated with densely packed crystallin proteins (1). In the absence of cellular machinery, crystallins are neither degraded nor regenerated. Therefore, maintaining their remarkable solubility throughout a lifetime is required for healthy lens function. Crystallin solubility can be reduced due to mutations, modifications, or changes in the lens environment to cause cataract, an opacification of the lens that causes blindness if untreated. Several known mutations are associated with hereditary cataract, e.g., in the genes for α-crystallin (2,3,4), γD-crystallin (5,6,7), and γS-crystallin (8,9). In contrast, aggregation in age-related cataract is often driven by post-translational modifications (PTMs). Various PTMs have been reported from patient samples, including truncation (10), deamidation (11,12), and disulfide bond formation (13).

Protein aggregation is central to a plethora of diseases. Amyloid fibrils are associated with neurodegenerative disorders, such as Alzheimer’s (14,15), Parkinson’s (16,17), and Huntington’s (18) diseases. Similar deposits in the pancreas disrupt insulin production and cause type 2 diabetes (19,20). These aggregates commonly have a cross-β structure that does not necessarily resemble the folded monomer; their formation requires permanent unfolding and subsequent misfolding. Although amyloid formation is a particularly well-known aggregation mechanism, many disease-associated proteins also aggregate in native-like states. For instance, the polymerization of hemoglobin S in sickle cell anemia progresses via surface interactions, without unfolding (21,22). A cataract-related variant of γD-crystallin was also found to retain its folded structure within amorphous aggregates (23). Other crystallin variants of clinical significance have also been reported to aggregate well below their thermal unfolding temperatures (24,25,26). The underlying mechanism of these observations is often presumed to be transient unfolding, which exposes buried hydrophobic surfaces as hotspots for undesired interactions. However, it is also possible that a modified protein surface leads to altered intermolecular interactions. Beyond cataract, crystallin proteins can serve as a useful model system for studying this alternative mode of aggregation.

Oxidative stress is implicated in age-related cataract (27,28), as well as many amyloidogenic diseases (29). Ultraviolet (UV) radiation is a major source of oxidative stress for crystallins (30), given the lens’s daily exposure to sunlight. In age-related cataract, UV-induced PTMs on crystallins accumulate, progressively increasing their susceptibility to further damage and eventual aggregation (31). A major modification caused by oxidative stress is deamidation of Asn and Gln side chains to form Asp and Glu, as well as products formed by racemization and isomerization (32,33,34,35,36,37). In aged lenses, this type of modification is more commonly observed in unstructured linkers rather than in the well-defined β sheet regions of the protein (38). A large body of work on structural crystallins has found that deamidation at some sites causes reduced stability and increased aggregation (39,40,41,42), as well as structural perturbations, including destabilization of the dimer structure in β-crystallins and alteration of its dynamics (43,44). In γS-crystallin, deamidation of surface Asn and Gln residues promotes aggregation (45,46,47), whereas in γD-crystallin, several single-site deamidations on the protein surface have minimal impact on structure and stability (48). Extensive γ-irradiation of human γS-crystallin resulted in little impact on solubility, despite the presence of multiple oxidation sites per protein (49). Overall, these observations suggest that crystallins are highly resistant to PTMs, such that even extensive modification often causes subtle changes to dynamics and interprotein interactions rather than large-scale misfolding or catastrophic aggregation.

The common hydrophobic core of vertebrate γ-crystallins has four highly conserved Trp residues that are thought to resist photodamage (50,51,52). The fluorescence of these residues is rapidly quenched, dissipating energy through Förster resonance energy transfer (FRET) and charge transfer before photochemical reactions can occur. A prior study of UV- and γ-irradiated γS-crystallin revealed several oxidation products. Notably, W163—but not any of the other three conserved Trp residues—was oxidized (49). This finding was consistent with earlier molecular dynamics (MD) simulations suggesting that W163 is the sole solvent-exposed Trp (53) and with an experimental study of γ-crystallins in γ-irradiated rat lenses (54). Mass spectrometry (MS) revealed the addition of a single oxygen atom at W163 in the γ-irradiated protein. We were unable to identify the exact position(s) of the modification; however, 5-hydroxytryptophan (5HTP) has been found in other oxidized proteins (55,56). Unlike deamidation (48,57), which has been more extensively studied, this side-chain modification cannot not be easily mimicked via mutagenesis.

Here, we use genetic code expansion (GCE) to achieve site-specific oxidation of the critical residue, W163. GCE is often used to expand chemical functionality by incorporating unnatural amino acids. This methodology has great potential for introducing oxidative damage at selected sites, enabling detailed biophysical and chemical characterization. For example, 3-nitrotyrosines were installed in α-synuclein (58) and indoleamine 2,3-dioxygenase 1 (59) to study the impact of this PTM on specific sites. We replaced W163 of γS-crystallin with 5HTP. Our objective is to investigate the structural and biophysical impacts of covalent oxidative damage on crystallins. GCE enables characterization of this post-translationally acquired oxidation using methods that require high-purity, recombinantly expressed, and isotopically labeled protein samples. We hypothesized that introducing a polar hydroxyl group into the hydrophobic core would disrupt the structural integrity of the C-terminal domain. Then, we compared wild-type γS-crystallin (γS-WT) with the variant in which 5HTP replaces Trp at position 163 (γS-W163(5HTP)). We find that although γS-W163(5HTP) displays minimal structural disruption at ambient temperatures, it is less resistant to both thermal and chemical denaturation and more susceptible to thermal aggregation.

Materials and methods

Plasmid propagation was done in Invitrogen Escherichia coli DH5α cells. Plasmids were isolated using a QIAprep Spin Miniprep Kit, and genes of interest within plasmid samples were sequenced using GENEWIZ Sanger Sequencing with the T7 Universal Primer. Expression of the γS-WT was done in Novagen E. coli Rosetta (DE3) cells. The pET28a(+) vector containing the codon-optimized γS-WT sequence with an N-terminal hexahistidine tag and a tobacco etch virus (TEV) protease cleavage sequence was previously purchased (57). An agar stab of ATMW-BL21 cells containing pEVOL suppression plasmid with glnS and lpp promoters (pEVOL-glnS-lpp) was gifted by Prof. A. Chatterjee (60) and used to express γS-W163(5HTP). The pET28a(+) vector containing the His-tagged γS-WT construct with an amber codon, TGA, at residue 163 was purchased from Twist Bioscience (South San Francisco, California). Corning disposable sterile bottle-top filters with a 0.45 μm membrane and Fisherbrand sterile syringe filters with 0.45 or 0.22 μm PVDF membranes were purchased from Fisher Scientific (Pittsburgh, Pennsylvania). Whatman Anotop syringe filters with a 0.02 μm membrane, HisTrap HP (Ni-IMAC), and HiLoad Superdex 75 pg (size-exclusion chromatography [SEC]) columns were purchased from Cytiva (Marlborough, Massachusetts). Spectra/Por dialysis tubing (6–8 kDa) was purchased from Repligen (Waltham, Massachusets), and the Amicon ultra-centrifugal filter (3 kDa) was purchased from MilliporeSigma (Burlington, Massachusetts).

Protein expression and purification

The following concentrations of antibiotics were used in growth media: 95 μg ml⁻1 spectinomycin, 20 μg ml⁻¹ chloramphenicol (Chlor), 15 μg ml⁻¹ Zeocin, 10 μg ml⁻¹ gentamycin, and 50 μg ml⁻¹ kanamycin (Kan). Unless specified otherwise, purification buffers contained 0–500 mM imidazole, 50 mM sodium phosphate, 300 mM sodium chloride, and 0.05% sodium azide in water (pH 7.0) and were sterilized by filtration using 0.45 μm filters and degassed under vacuum.

Expression of γS-WT

Expression of γS-WT was adapted from previously described protocols (53). Plasmid containing the γS-WT sequence was propagated and transformed, and a single colony was grown in Luria-Bertani (LB) broth (1% tryptone, 0.5% yeast extract, and 1% sodium chloride in water) supplemented with chloramphenicol and kanomycin overnight (14–16 h) at 37°C with shaking. Overexpression was achieved using autoinduction media inspired by Studier’s ZYM-5052 media (61): autoclaved media containing final concentrations of 1% tryptone and 0.5% yeast extract in water was supplemented with stock solutions sterilized by filtration to final concentrations of 50 mM sodium phosphate, 50 mM potassium phosphate, 25 mM ammonium sulfate, 0.05% glucose, 0.5% lactose, 0.6% glycerol, and 2 mM magnesium sulfate. Overnight culture was diluted 1:20 into 500 ml of autoinduction media supplemented with Chlor and Kan for 1 h at 37°C followed by 24–28 h at 25°C with shaking.

Expression of γS-W163(5HTP)

Chemically competent ATMW-BL21 pEVOL-glnS-lpp cells were transformed with a plasmid encoding for W163TGA. A single colony was grown in LB broth supplemented with spectinomycin, chlor, Zeocin, gentamycin, and kanomycin overnight at 37°C with shaking. Overnight culture was diluted 1:20 into 500 ml of LB broth supplemented with antibiotics and a final concentration of ±4 mM 5HTP. Cells were allowed to grow at 37°C until an OD₆₀₀ of 0.8–1.0 was reached. Overexpression was induced by adding a final concentration of 1 mM IPTG and incubating for 20–24 h at 25°C with shaking.

¹⁵N-labeled expression

¹⁵N-labeled expression was adapted from previously described protocols (62). A single colony of γS-WT- or γS-W163(5HTP)-expressing bacteria was grown in LB broth supplemented with corresponding antibiotics overnight at 37°C with shaking. Overnight culture was diluted 1:20 into 500 ml of LB broth supplemented with corresponding antibiotics, and cells were allowed to grow at 37°C until an OD₆₀₀ of 1.0 was reached. Cells were harvested by centrifugation at 4000 rpm at 4°C for 20 min and resuspended in 500 ml of ¹⁵N-labeled media (48 mM sodium phosphate, 22 mM potassium phosphate, 9 mM sodium chloride, 2 mM magnesium sulfate, 0.1 mM calcium chloride, 18 mM ¹⁵N-labeled ammonium chloride, and 55 mM glucose in water [pH 7.0]) supplemented with corresponding antibiotics and 5HTP as necessary. Cells were allowed to grow at 37°C for 1 h, and then overexpression was induced by adding a final concentration of 1 mM IPTG and incubating for 40–44 h at 25°C with shaking.

Purification

Upon overexpression, cells were harvested by centrifugation at 4000 rpm at 4°C for 30 min, flash-frozen in liquid nitrogen, and then stored at −80°C. Frozen cells were resuspended 10 ml of lysis buffer (25 mM imidazole) per gram of pellet. Resuspended cells were supplemented with 10 mM magnesium sulfate, 1 μg ml⁻¹ lysozyme, 0.1 μg ml⁻¹ deoxyribonuclease I, and 0.7 μL ml⁻¹ β-mercaptoethanol, incubated on ice for 30 min, and then lysed by sonication (2 s on/off, 15 min, 40% amplitude). The resulting lysate was clarified by centrifugation at 13,000 rpm at 4°C for 45 min, and the supernatant was loaded onto a Ni-IMAC column. A gradient of 25–500 mM imidazole was used to elute the His-tagged protein, and then 1 mg of Loening’s TEV Hexa (prepared in-house) was added per 150 mg of protein to cleave the His tag (63). The resulting mixture in hydrated dialysis tubing was placed in dialysis buffer (50 mM sodium phosphate and 150 mM sodium chloride in water [pH 7.0]) and stirred at 4°C for 16 h. Reapplication to the Ni-IMAC column removed the TEV protease, and the cleaved protein was reduced with a final concentration of 1 mM dithiothreitol (DTT), concentrated, and loaded onto a SEC column. The desired protein was eluted using SEC buffer (0 mM imidazole), dialyzed in water, concentrated and aliquoted, and then lyophilized for storage at −80°C.

MS

Intact protein samples were analyzed by liquid chromatography-MS (Waters Xevo G2-XS QTof) equipped with a phenyl column (ACQUITY UPLC BEH Phenyl VanGuard Column). A gradient of 0–90% buffer B (acetonitrile) in buffer A (0.1% formic acid in water) at a 0.2 ml/min flow rate for 0.5 min was used to separate proteins from buffer salts. Electronspray ionization (ESI) was operated in positive mode (400–4000 Da) with a capillary voltage of 3000 V and a cone voltage of 40 V. Nitrogen was used as desolvation gas at 350°C and a total flow of 800 l h⁻¹. Total average mass spectra were reconstructed from the charge state series using the MaxEnt1 algorithm from MassLynx software.

Pepsin digest and tandem MS

Protein samples were precipitated in precipitation buffer (100 mM hydrochloric acid and 25 mM zinc sulfate in acetonitrile) and pelleted by centrifugation. Samples were decanted to remove the supernatant, and the pellets were resuspended in alkylation buffer (250 mM ammonium bicarbonate, 50 mM ethylenediaminetetraacetic acid, and 67 mM DTT in water [pH 8.0]) and incubated at 80°C for 10 min. Alkylation was carried out at 80°C for 10 min by supplementing the mixture with a final concentration of 118 mM acrylamide and was quenched by bringing the concentration of DTT up to 200 mM and incubating at 80°C for 10 min. Reaction mixtures were diluted 1:10 in precipitation buffer, and then protein samples were pelleted by centrifugation. Samples were decanted to remove the supernatant, and the pellets were resuspended in pepsin solution (25 mg/l pepsin, 25 mM hydrochloric acid, and 100 mM potassium chloride in water). Protein samples were digested overnight at 37°C, and the reaction was quenched by adding a final concentration of 91 mM sodium hydroxide and incubating for 5 min. Samples were re-acidified using hydrochloric acid and diluted in 1% formic acid in water for analysis.

Digested protein samples were analyzed by liquid chromatography-tandem MS (Waters Xevo G2-XS QTof) equipped with a phenyl-hexyl column (ACQUITY Premier CSH Phenyl-Hexyl Column). A gradient of buffer B (acetonitrile) in buffer A (0.1% formic acid in water), as shown in Table 1, at a 0.4 ml/min flow rate was used to separate peptide fragments. Cycles of scans were collected throughout the retention time using fast data-directed analysis. A single 0.2 s MS1 scan was acquired to identify multiply charged peptide precursors, followed by two 0.4 s MS2 scans targeting the two peaks of highest intensity, applying a collision energy ramp of 20–40 V. Raw data files were exported, converted to mzML files using MSConvert, and imported to PEAKS Studio 12. Areas under peaks corresponding to peptide fragments containing Trp or oxidized Trp were tabulated across samples.

Table 1.

Buffer gradient used for liquid chromatography-MS/MS for digested protein samples

Time (min)	Buffer A (%)	Buffer B (%)
0	100	0
1	100	0
23	70	30
25	0	100
27	0	100
28	100	0
30	100	0

Biophysical characterization

Lyophilized protein samples were reconstituted in phosphate buffer (10 mM sodium phosphate in water [pH 7.0]) that was sterilized by filtration and degassed under vacuum, then reduced with a final concentration of 1 mM DTT. After incubating on ice for 1 h, samples were filtered using 0.02 μm filters, analyzed using a NanoDrop One spectrophotometer for A₂₈₀, and then diluted to the desired concentration in phosphate buffer, unless specified otherwise. Concentration was calculated from the absorbance at 280 nm using the extinction coefficient for γS-WT, 41,040 M⁻¹ cm⁻¹ (64). Statistical details are reported in the figure legends and the method details, and data are presented as the mean ± standard deviation (n = 3).

Absorbance and fluorescence

Absorbance between 200 and 800 nm of protein samples (1 mg/ml) was measured at ambient temperature using a Cary-60 Absorption Spectrometer. λ_max of absorbance was used (5 nm slit widths) to scan for fluorescence between 300 and 500 nm at ambient temperature using a Cary Eclipse Fluorimeter. The corresponding λ_max of emission was used to scan for the λ_max of excitation between 250 and 300 nm, which was then used to produce the reported emission spectrum.

Circular dichroism and thermal unfolding

Molar ellipticity of circularly polarized light between 190 and 260 nm of protein samples (0.1 mg/ml) was measured at ambient temperature using a Jasco J-810 Spectropolarimeter. Molar ellipticity at 218 nm of protein samples (0.25 mg/ml, 10 mM sodium phosphate, and 150 mM sodium chloride in water [pH 7.0]) was measured as a function of temperature as previously described (65). Molar ellipticity was converted to fraction unfolded by setting the minimum value as the fully folded state and the maximum value as the fully unfolded state within each sample. Thermal unfolding data were fitted to a nonlinear regression model using a 4-parameter logistic curve in GraphPad Prism software. The 4-parameter model, which is a common approach to fitting this type of data, describes a sigmoidal curve with a midpoint, slope, and lower and upper plateaux (66).

Chemical unfolding and thermodynamics

Intrinsic tryptophan fluorescence of protein samples was measured using a TECAN Spark Multimode Microplate Reader. Sample preparation was adapted from a previously described protocol (65). Protein samples (0.1 mg/ml, 0–6 M guanidine hydrochloride in 0.2 M increments, 10 mM sodium phosphate, and 50 mM sodium chloride in water [pH 7.0]) were incubated at ambient temperature for 48 h with shaking and then excited at 260 nm (5 nm slit widths), and emission intensity was recorded between 300 and 400 nm. Chemical unfolding was assessed by calculating the fluorescence ratio of 355 to 325 nm and then converting it to fraction unfolded by setting the minimum value as the fully folded state and the maximum value as the fully unfolded state within each sample. Thermodynamic parameters were calculated assuming a two-state unfolding mechanism as previously described (65,67).

Dynamic light scattering

Aggregation of protein samples (1 mg/ml) was assessed via dynamic light scattering using a Malvern Zetasizer Nano-ZS as a function of temperature. Average particle sizes were measured across a temperature range of 25°C–80°C in 1°C increments.

MD

γS-crystallin models were created from PDB: 2M3T (68); two monomers were placed in a cubic cell of ≈92.5 Å in length, at a concentration of ≈50 mg/ml. The cell was filled with TIP3P water (69), counterions were added to neutralize the structure, and additional sodium chloride ions were added to a total ion concentration of 10 mM. Residue protonation states were calibrated to pH 7 using PROPKA3 (70), and the combined system was prepared using VMD (71). Four trajectories were simulated, representing γS-WT or γS-W163(5HTP) at 37°C or 78°C, respectively. Trajectory simulation was performed using NAMD (72) with the CHARMM36m force field (73) under periodic boundary conditions in an NpT ensemble. Force field parameters for the 5HTP side chain were obtained from Biomolecular Modeling Database entry BMOD0000009519 (74). Nosé-Hoover Langevin piston pressure control (75,76) was used to achieve 1 atm of pressure, and Langevin temperature control with a damping coefficient of 1/ps was applied. After an initial 25 ns equilibration phase, each system was simulated for 100 ns, during which 5000 frames were retained for analysis (1/20 ps).

Trajectory analysis was performed using the bio3d (77), sna (78), and network (79) libraries for R (80). Each full trajectory was downsampled to 1000 frames (1/100 ps), and the periodic cell was unfolded to an infinite repeating population. Each frame was then examined to determine whether the proteins were in a monomeric, dimeric, or oligomeric state. Binding was determined by van der Waals contact. Visualizations of representative states were obtained using the RGL library. (81).

¹H-¹⁵N HSQC NMR

¹⁵N-labeled protein samples (a final concentration of 0.5 mM; 10 mM sodium phosphate in water supplemented with 10% deuterium oxide and 0.5 mM sodium trimethylsilyl propanesulfonate [pH 7.0]) were characterized as a function of temperature (25°C–60°C for γS-WT and 25°C–55°C for γS-W163(5HTP) in 5°C increments). ¹H-¹⁵N heteronuclear single-quantum coherence (HSQC) (hsqcfpf3gpphwg) experiments were performed on a Bruker Avance Neo 800 MHz. Samples were equilibrated for over 10 min at each temperature before data acquisition. ¹H shifts were referenced to the internal standard, 10% deuterium oxide and 0.5 mM sodium trimethylsilyl propanesulfonate, and ¹⁵N shifts were referenced indirectly. The resulting spectra were processed with Bruker TopSpin software, and peak assignments and chemical shift perturbation (CSP) calculations were performed using CcpNmr Analysis v.3 (82). ¹H shifts as a function of temperature were plotted and fitted to a linear regression model in Graphpad Prism software to calculate amide proton temperature coefficients.

Results

Site-specific incorporation of 5HTP in place of W163 yields minimal changes to the overall structure

Our model oxidation product, 5HTP, was encoded at residue 163 of γS-crystallin using an amber codon (TGA) and incorporated via an engineered E. coli strain developed by Ficaretta et al. (60). Data confirming successful expression and purification with the site-specific incorporation of 5HTP are shown in Figs. S1 and S2 and Tables S1 and S2. The structure of γS-W163(5HTP) at room temperature was investigated using circular dichroism (CD) spectroscopy (Fig. 1 A) and intrinsic Trp fluorescence (Fig. 1 B). CD spectroscopy reports on protein secondary structure by measuring the difference in absorption between left-handed and right-handed polarized light; γ-crystallins have primarily β strand secondary structure, which is characterized by a strong negative signal at 218 nm (84). The CD spectrum of γS-W163(5HTP) is nearly identical to γS-WT, indicating that the β strand structure of this protein’s double-Greek key architecture is nearly intact; however, the characteristic minimum at 218 nm is slightly shallower in γS-W163(5HTP), suggesting only subtle structural differences.

Figure 1.

γS-WT and γS-W163(5HTP) were similarly folded at ambient temperature. (A) CD spectra prominently indicate a β sheet structure for both proteins. Slightly less pronounced molar ellipticity was observed for γS-W163(5HTP), suggesting subtle changes in conformation. (B) Normalized fluorescence spectra of the intrinsic tryptophan fluorescence showed a slight narrowing of the emission peak upon oxidation, as expected with the change in chromophore structure (83).

In pursuit of enhanced sensitivity for structural disruption caused by the modification, we used intrinsic Trp fluorescence, which is capable of reporting on the local environment and folding status of the hydrophobic core in γ-crystallins. Local unfolding that increases hydration around any of the Trp resonances partially quenches this fluorescence, leading to red shifting of the emission maximum (85). The emission peak of γS-WT matched that of previous reports; its broad profile is thought to be an overlap of the emission peak of the N-terminal FRET acceptor (λ_em = 331 nm) and that of the C-terminal FRET acceptor (λ_em = 324 nm) (51). For γS-W163(5HTP), we observe a minor shift in the emission maximum (λ_em) from 328 to 330 nm, along with a slight narrowing of the spectrum. Given that the emission of 5HTP within folded proteins is generally lower in energy (83), this result is consistent with the installed hydroxyl group to the C-terminal acceptor resulting in a better overlap between the two FRET pairs, most likely in the absence of unfolding around it. Taken together, these results strongly suggest that the incorporation of 5HTP at position 163 has only minor effects on the structure of γS-crystallin.

Oxidation of W163 reduces thermal and chemical stability and increases aggregation propensity

The spectral data for chemical and thermal unfolding of γS-WT and γS-W163(5HTP) are shown in Fig. 2 A–C, and the quantitative data are tabulated in Table 2. Chemical denaturation was monitored by the redshift in the intrinsic tryptophan fluorescence upon unfolding after incubation at increasing concentrations of guanidine hydrochloride (Gdn-HCl) (Fig. 2 A). The midpoint unfolding concentration is in reasonable agreement with a prior measurement of 2.6 M for γS-WT (46). Both γS-WT and γS-W163(5HTP) produced a largely sigmoidal curve, suggesting that the two-state unfolding mechanism previously assumed for γS-WT (65) was also extendable to γS-W163(5HTP). The ratio of unfolded to folded conformations at each equilibrium was used to calculate the ΔG of denaturation (ΔG_app) (Fig. 2 B). At their respective midpoint [Gdn-HCl] values, the ΔG_app of γS-W163(5HTP) is about 0.5 kJ/mol lower than that of γS-WT. Fig. 2 C shows the thermal unfolding of both proteins measured by the loss of β sheet structure detected via CD spectroscopy. The midpoint unfolding temperature (T_m) of γS-WT is 78.1°C, which is in reasonable agreement with previously measured values of 75°C (86) and 77°C (57); this is reduced to 72.8°C for γS-W163(5HTP).

Figure 2.

Decreased stability and increased aggregation propensity were observed for the modified protein. Points indicate mean values; error bars denote standard deviation. (A) γS-W163(5HTP) unfolded at lower concentrations of the denaturant Gdn-HCl compared to γS-WT. (B) Chemical unfolding curves of both proteins emulated two-state unfolding, yielding linear relationships between ΔG_app and [Gdn-HCl]. ΔG_app was calculated assuming a two-state unfolding mechanism, using K = e − Δ^Gapp/RT (67). (C) Thermal unfolding curves were generated by observing the molar ellipticity at 218 nm as a function of temperature. γS-W163(5HTP) exhibited similar unfolding behavior to γS-WT but at lower temperatures. (D) Growth in average particle size under increasing temperatures was studied using dynamic light scattering. A set of representative data were plotted for γS-W163(5HTP) with 95% confidence intervals calculated from three sample replicates shown in dotted lines.

Table 2.

Destabilization of γS-W163(5HTP) measured against γS-WT

Parameter	γS-WT	γS-W163(5HTP)	p value
Tm (°C)	78.1 ± 0.4	72.8 ± 0.3	8.89 × 10−5	∗∗∗
Midpoint (M)	2.77 ± 0.08	2.47 ± 0.02	0.018	∗
ΔGapp (kJ/mol)	6.01 ± 0.03	5.5 ± 0.8	0.42	–
m (kJ/mol/M)	−0.75 ± 0.03	−0.7 ± 0.1	0.48	–

Unfolding midpoint temperatures (T_m) were derived from thermal unfolding curves, and thermodynamic parameters were calculated from chemical unfolding in Gdn-HCl. Parameters are given as the means ± one standard deviation; p values are from two-sided Welch tests of equality between variants. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

Thermal aggregation propensity was measured using dynamic light scattering, which reports the apparent particle size. A well-folded monomer of γS-crystallin has a hydrodynamic radius of 2.5 nm (46). The apparent particle size as a function of temperature is shown for γS-WT and γS-W163(5HTP) in Fig. 2 D. As reported in the literature, γS-WT oligomers exhibited a gradual increase in size before reaching its aggregation temperature, T_agg, at about 43°C. In contrast, γS-W163(5HTP) formed large oligomers well below physiological temperatures, starting at about 34°C. Collectively, these results indicate that although γS-W163(5HTP) is only slightly less stable than γS-WT, it is markedly more aggregation-prone at temperatures typical of the native lens environment.

MD suggests a mechanism of differential aggregation behavior

The difference in γS-WT versus γS-W163(5HTP) aggregation behavior below their respective unfolding temperatures raises the question of how aggregation occurs before unfolding. To investigate this question, we ran atomistic MD simulations of γS-crystallin γS-WT and γS-W163(5HTP) at 37°C and 78°C. Simulation was performed under periodic boundary conditions at ≈50 mg/ml, with two simulated monomers within the periodic cell. Cell conditions were designed to allow for unbound monomeric, bound dimeric, or higher-order oligomeric states. Fig. 3 B illustrates the respective states. As shown in Fig. 3 A, both variants are predominantly dimeric at 37°C. On the MD timescale (100 ns), no unfolding was observed at 78°C for either protein, enabling probing of interactions at increased temperatures in the folded state. For γS-WT, heating leads to an increase in the monomeric population; this is compatible with a “sticky rigid ball” model, in which monomers are weakly attracted but increasingly shaken apart at moderate temperatures. By contrast, γS-W163(5HTP) becomes almost entirely oligomeric at 78°C. This appears to result from increasing favorability of “end-to-end” binding in γS-W163(5HTP) (Fig. 3 B) rather than binding either near or below the interdomain cleft. Oxidation of W163 thus appears to mildly “soften” the structure of γS-crystallin in a manner that enhances binding to the edges of the respective domains.

Figure 3.

MD shows that heating has distinct effects on γS-WT and γS-W163(5HTP) before unfolding. (A) Respective monomeric, dimeric, and oligomeric state occupancies at 37°C and 78°C, at ≈50 mg/ml. Both variants are predominantly dimeric at 37°C; heating to 78°C increases free monomers in γS-WT, whereas this induces oligomerization in γS-W163(5HTP). (B) Typical examples of monomeric, dimeric, and oligomeric conformations; the oligomeric state is of indefinite size due to periodic boundary conditions.

Solution-state NMR reveals local structural changes caused by oxidation of W163

We used solution-state NMR spectroscopy to gain residue-level insights into what structural changes are induced upon oxidation of W163, particularly under the conditions of observed and simulated thermal aggregation. ¹H-¹⁵N HSQC experiments are an established method to track the movements of the backbone amides in a polypeptide chain, as their chemical shifts are highly sensitive to local conformational changes. Using the resonance assignments available for γS-WT (62), the HSQC spectrum of γS-W163(5HTP) taken at 25°C was assigned by analogy (Fig. S3). The Euclidean distances between each pair of resonances on superimposed spectra were plotted (Fig. 4 B). The μ and 1–4 σ of these CSP values were then used to color the solution-state structure of human γS-crystallin (PDB: 2M3T) (68) according to the legend in Fig. 4 A. As expected, almost no deviations are seen for the residues in the N-terminal domain, whereas some variations are observed in the C-terminal domain, particularly around W163. Even the neighboring residues, however, exhibited CSP values of less than 0.1 ppm, which corroborated our earlier findings of γS-W163(5HTP) retaining γS-WT-like structure and stability.

Figure 4.

Oxidation at W163 causes subtle structural changes to the C-terminal domain of human γS-crystallin. (A) Structure of human γS-crystallin (PDB: 2M3T) (68). Residues with chemical shift perturbation (CSP) values within one standard deviation of the mean are colored in lavender, and residues with CSP values within two, three, or four standard deviations of the mean are highlighted in progressively redder shades. The indole side chain of W163 is also shown in the corresponding color based on its CSP value. Residues that were not detected or assigned on both spectra are shown in dark gray. (B) CSP values were calculated using ¹H-¹⁵N HSQC NMR spectra of γS-WT and γS-W163(5HTP) at 25°C. Dotted lines indicate the mean (lavender) plus one (dark salmon), two (bright coral), or three (brick red) standard deviation(s).

To further investigate the nuanced increase in structural fluidity upon oxidation of W163 suggested by MD simulations, we monitored how the proton chemical shifts of both proteins changed as a function of temperature (Fig. S4). Residues that are hydrogen bonded to the solvent, as opposed to other residues within the same molecule, typically display a large negative slope when their proton chemical shifts are plotted versus temperature (87). An amide proton temperature coefficient more negative than −4.6 ppb/K is often interpreted as the absence of intramolecular hydrogen bonds at the residue. As expected for a folded, globular protein, γS-crystallin exhibits a range of temperature coefficients, consistent with a mixture of solvent-exposed and intramolecularly hydrogen-bonded amide protons (Fig. S5). Most residues in γS-W163(5HTP) yield similar values; only two residues show a difference that crosses the −4.6 ppb/K threshold. One residue is involved in an intramolecular hydrogen bond in γS-WT but interacts with the solvent in γS-W163(5HTP): R158 (γS-WT: 0.02; γS-W163(5HTP): −7.22). A166 also shows a drastic perturbation, but it moves in the opposite direction (γS-WT: −10.87; γS-W163(5HTP): −2.816). Both residues that undergo changes in H-bonding status are located within a few residues of the installed oxidation mimic. Such rearrangement of the H-bonding network that stabilizes the local secondary structure may cause transient unfolding that is difficult to capture experimentally but is critical for the observed premature thermal aggregation of γS-W163(5HTP).

Discussion

A common theme observed in studies of cataract-related crystallin variants is that mutations and PTMs that have only minimal impact on the structure at room temperature nevertheless result in diminished stability and/or increased aggregation propensity. The G18V variant of γS-crystallin associated with childhood-onset cataract, for instance, does not exhibit large-scale unfolding or misfolding but instead shows dramatically lower thermal stability and solubility (86,88). Other examples include the P23T (γD) (89), R168W (γC) (90), and R10P (γA) (91) variants of γ-crystallins. A similar trend is observed with a number of deamidation mimics: minimal structural deviations from γS-WT are seen under ambient and buffered conditions using crystallographic and spectroscopic methods (46,47,57). Nonetheless, they exhibit cumulative destabilization and a higher propensity for polymerization under stress. It is noteworthy that nine sites of deamidation added up to a 5.1°C decrease in T_m as previously reported by our laboratory, whereas the single-site, single-atom modification in γS-W163(5HTP) resulted in a 5.3°C decrease. This comparison reinforces the significance of Trp residues in maintaining the structural integrity of γS-WT, particularly its tightly folded hydrophobic core. Introducing an electronegative atom onto the nonpolar face of well-stacked β sheets compromises thermal stability more than installing multiple surface-charge modifications, although both types of PTM are structurally tolerated in the absence of heat.

Deamidated γS-crystallin variants also show an increase in intramolecular disulfide bonding (57); the formation of disulfide-bonded dimers are themselves stable but may represent intermediates along the path to extensive aggregation (45,92,93). The N76D variant studied by Ray et al. exemplifies this “stability paradox,” in which its dimeric form demonstrates enhanced thermal stability and initial resistance to thermal aggregation. Although dimerization provides a temporary shield against unfolding, disulfide-linked dimers aggregate faster to produce higher-molecular-weight particles via non-cooperative unfolding. On the contrary, MD simulations of γS-W163(5HTP) at elevated temperatures displayed folded monomers that are vulnerable to subtle changes in intermolecular surface interaction. Simulated trajectories suggest the formation of extended aggregates, possibly through the development of “sticky” patches on the outer portion of each domain, away from the central cleft where intermolecular disulfide bond formation occurs. Experimentally, the thermal aggregation of γS-W163(5HTP) produced a distinct profile than those of γS-WT and the aforementioned deamidation variants, which are roughly sigmoidal. The oxidized variant formed oligomers even before reaching physiological temperature, growing exponentially in size at every increment. Collectively, these data suggest that an alternative mode of aggregation with a distinct mechanism of intermolecular attraction is likely in effect in γS-W163(5HTP). Characterizing the proposed regions of altered surface interaction would seem fruitful for further investigation. In particular, investigating the dynamics of the residues neighboring the installed oxidation mimic located at the C-terminal end of the protein, with a focus on those whose H-bonding interactions are displaced, is a promising future direction. Another key objective for future work is to measure the second virial coefficient of γS-W163(5HTP). This experimental observable provides a direct measurement of intermolecular attraction between particles (94), which will be highly useful for determining the mechanism of aggregation.

Data and code availability

• •All data are available at Mendeley data: https://doi.org/10.17632/3krz2d92x8.1.
• •No original code was generated for this study.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We acknowledge D. Fishman and the UCI Laser Spectroscopy Labs, Ben Katz and Felix Grün of the UCI Mass Spectrometry Facility, and Suvrajit Sengupta of the UCI NMR Facility for access to instrumentation, assistance with data collection, and helpful discussions. This work was supported by the National Institutes of Health under award numbers R01GM144964 to C.T.B. and R.W.M. and R01EY021514 to R.W.M.

Author contributions

Y.S., Z.G.L., and R.W.M. planned and designed the experiments. Y.S., Z.G.L., T.K.D., and A.A.A. performed the experiments. C.T.B. designed and performed simulation modeling. Y.S., Z.G.L., C.T.B., and R.W.M. performed data analysis. Figures were prepared by Y.S., C.T.B., and R.W.M. The manuscript was written and edited by Y.S., C.T.B., and R.W.M. All authors approved the final version of the manuscript.

Declaration of interests

The authors declare no competing interests.